whisper-medium-swagen-combined-10hrs-model

This model is a fine-tuned version of openai/whisper-medium on the swagen dataset. It achieves the following results on the evaluation set:

• Loss: 0.4674
• Wer: 0.3075

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.6418	0.2385	200	0.8312	0.4835
1.9381	0.4769	400	0.6574	0.4050
1.7528	0.7154	600	0.5706	0.3545
1.7421	0.9538	800	0.5140	0.3504
0.9259	1.1931	1000	0.5175	0.3262
0.8254	1.4316	1200	0.4978	0.3096
0.9616	1.6700	1400	0.4755	0.2999
0.7893	1.9085	1600	0.4674	0.3075
0.3547	2.1478	1800	0.4848	0.3158
0.3568	2.3863	2000	0.4913	0.2616
0.3759	2.6247	2200	0.4685	0.3104

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0